Oncogene (2014), 1–10 © 2014 Macmillan Publishers Limited All rights reserved 0950-9232/14 www.nature.com/onc

ORIGINAL ARTICLE

Intratumoral heterogeneity of ADAM23 promotes tumor growth and metastasis through LGI4 and nitric oxide signals ET Costa1,2, GF Barnabé1,2, M Li3, AAM Dias4, TR Machado2, PF Asprino1,2, FP Cavalher2, EN Ferreira5, M del Mar Inda3, MH Nagai6, B Malnic6, ML Duarte1,2, KRM Leite7, ACSD de Barros8, DM Carraro5, R Chammas7, HA Armelin6,9, W Cavenee3, F Furnari3 and AA Camargo1,2 Intratumoral heterogeneity (ITH) represents an obstacle for cancer diagnosis and treatment, but little is known about its functional role in cancer progression. The A Desintegrin And Metalloproteinase 23 (ADAM23) gene is epigenetically silenced in different types of tumors, and silencing is often associated with advanced disease and metastasis. Here, we show that invasive breast tumors exhibit significant ADAM23-ITH and that this heterogeneity is critical for tumor growth and metastasis. We demonstrate that while loss of ADAM23 expression enhances invasion, it causes a severe proliferative deficiency and is not itself sufficient to trigger metastasis. Rather, we observed that, in ADAM23-heterotypic environments, ADAM23-negative cells promote tumor growth and metastasis by enhancing the proliferation and invasion of adjacent A23-positive cells through the production of LGI4 (Leucine-rich Glioma Inactivated 4) and nitric oxide (NO). Ablation of LGI4 and NO in A23-negative cells significantly attenuates A23-positive cell proliferation and invasion. Our work denotes a driving role of ADAM23-ITH during disease progression, shifting the malignant phenotype from the cellular to the tissue level. Our findings also provide insights for therapeutic intervention, enforcing the need to ascertain ITH to improve cancer diagnosis and therapy. Oncogene advance online publication, 24 March 2014; doi:10.1038/onc.2014.70

INTRODUCTION Cancer is a dynamic heterogeneous disease. The coexistence of tumor cells with different phenotypic traits within a primary tumor has been known since the late 1970s. This phenotypic diversity, called intratumoral heterogeneity (ITH), appears to be the end result of branched evolutionary tumor growth, fostered by increased genetic and epigenetic instability and selective pressures from the tumor environment.1–8 Recent studies, mostly using deep-sequencing technologies, have elucidated the genetic basis of ITH in different types of tumors and associated metastatic sites.9–19 Collectively, these studies have demonstrated that ITH is a ubiquitous characteristic of neoplasms, arising during disease progression, and that primary tumors are ecosystems of evolving clones with different spatial and temporal distributions.20–22 Although ITH has been an underlying concept in cancer biology for decades,1–8,23 it is not yet fully considered by clinicians and pathologists and represents a major obstacle for cancer diagnosis and treatment. This is due mainly to technical limitations in assessing ITH in the clinical setting and to our rudimentary knowledge of the functional role of ITH for cancer progression, drug resistance and metastasis. The ADAM23 gene (also known as MDC3), encoding a non-catalytically active member of A Disintegrin And Metalloproteinase family,24 is frequently silenced in gliomas,25 breast,26,27 pancreatic,28 gastric,29 head and neck,30 colorectal31 and

lung32 tumors. We have previously demonstrated that ADAM23 expression is silenced by DNA promoter hypermethylation in the advanced stages of breast and head and neck tumors.26,27,30 In breast cancer, we showed that ADAM23 hypermethylation is an independent prognostic factor associated with a ninefold higher risk of developing distant metastases and lower survival rates.27 In addition, we demonstrated that knockdown of ADAM23 expression in the MDA-MB-435 cell line enhanced tumor cell migration, adhesion and arrest in the lungs of immunodeficient mice.27 Here, we show for the first time that loss of ADAM23 expression, previously associated with the development of distant metastases and poor prognosis, causes a proliferative and colonization deficiency and, counterintuitively, is not sufficient to trigger metastatic dissemination. Rather, the rise of ADAM23-negative (A23neg) cells within an ADAM23-positive tumor generates ADAM23-ITH and actively enhances proliferation and promotes invasion of adjacent ADAM23-positive (A23pos) tumor cells through Leucine-rich Glioma Inactivated 4 (LGI4) secretion and nitric oxide (NO) production, respectively. Ablation of LGI4 and NO production uncouples this cellular crosstalk and attenuates tumor cell proliferation and invasion. Our findings highlight a driving role of ADAM23-ITH in cancer progression by promoting tumor growth and metastasis. Our results also provide important insights for therapeutic intervention and enforce the need to address tumor subclonal architecture to improve the diagnosis and treatment of cancer.

1 Centro de Oncologia Molecular, Hospital Sírio Libanês, São Paulo, Brazil; 2Ludwig Institute for Cancer Research (LICR), São Paulo, Brazil; 3Ludwig Institute for Cancer Research (LICR), University of California, San Diego, CA, USA; 4Departamento de Biologia Geral (ICB), Universidade Federal de Minas Gerais, Belo Horizonte, Brazil; 5Centro Internacional de Pesquisa, Hospital AC Camargo, São Paulo, Brazil; 6Departamento de Bioquímica (IQ), Universidade de São Paulo, São Paulo, Brazil; 7Departamento de Urologia, Faculdade de Medicina, Universidade de São Paulo, São Paulo, Brazil; 8Departamento de Mastologia, Hospital Sírio Libanês, São Paulo, Brazil and 9Instituto Butantan, São Paulo, Brazil. Correspondence: Dr AA Camargo, Ludwig Institute for Cancer Research (LICR), Molecular Oncology Center, Hospital Sírio Libanês, Rua Cel. Nicolau dos Santos 69, São Paulo 01308-060, Brazil. E-mail: [email protected] or [email protected] Received 12 July 2013; revised 6 January 2014; accepted 14 January 2014

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

2

RESULTS ADAM23 silencing and ITH in invasive breast carcinomas To explore the role of ADAM23 silencing during breast cancer progression, we first analyzed if there were differences in ADAM23 mRNA expression levels between ductal breast carcinomas in situ (DCIS) and invasive ductal breast carcinomas (IDCs). Compared with normal breast, ADAM23 levels were reduced in most IDC (10/19), but not in DCIS (1/7) (Figure 1a, Po 0.05). We also analyzed the ADAM23 mRNA expression levels in paired in situ and invasive components of four IDCs to determine if ADAM23 silencing is homogeneous within IDC. Tumor cells from both components were laser microdissected (Figure 1b) and the heterogeneous ADAM23 expression was observed in two out of the four tumor samples analyzed (IDC2 and IDC4), most pronounced in tumor IDC4, extracted from a patient with extensive lymph node metastasis (Figure 1c). ADAM23 protein expression was also analyzed in 12 independent IDCs. As illustrated in Figure 1d, ADAM23 protein staining was more prominently detected in breast ducts with preserved normal architecture and a gradual decrease in staining intensity was observed between the in situ and invasive components of different IDCs (Figure 1d). It is worth noting that ADAM23-ITH was observed in topographically distinct invasive areas within undifferentiated IDCs, with some invasive areas being composed by a mosaic cluster of A23pos and A23neg tumor cells (Figure 1e, in detail, and Supplementary Figure 1).

These results indicate that ADAM23 expression is downregulated during breast cancer progression and that ADAM23ITH is a common characteristic of invasive carcinomas.

ADAM23-ITH promotes tumor cell proliferation and tumor growth To address the functional relevance of ADAM23-ITH during cancer progression, we first examined the role of ADAM23-ITH in tumor cell proliferation using three-dimensional cultures, which more effectively mimic tumor behavior,33 as well as by subcutaneous tumorigenesis assays. For these experiments, we used the MDA-MB-435 cell line, which we have previously used to address the functional impact of ADAM23 silencing in tumor cell adhesion and migration.27 In vitro, MDA-MB-435 ADAM23-positive cells (435-A23pos) or their ADAM23 short hairpin RNA (shRNA) derivatives (435-A23neg) (Supplementary Figures 2a and b) and a mixture of 1:1 ratio of both (435-A23het) were cultured as multicellular spheroids (MS), and proliferation rates were estimated by bromodeoxyuridine (BrdU) labeling. A lower percentage of BrdU-labeled cells was observed in 435-A23neg-MS (21%) compared with 435-A23pos-MS (32%) and 435-A23het-MS (32%, P o 0.01) (Figure 2a). In agreement with the proliferative deficiency of 435-A23neg cells, using anchorage-independent clonogenic assays, we observed that 435-A23neg cells formed an equivalent number of colonies to 435-A23pos cells, although these colonies were on average threefold smaller, confirming that both cells have a similar

Figure 1. ADAM23 expression levels in ductal breast carcinomas. (a) RT–qPCR analysis of ADAM23 mRNA expression in normal breast (NB) (n = 4), DCIS (n = 7) and IDCs (n = 19). (b) Representative example of a laser capture microdissection of the paired in situ (top panels) and invasive (bottom panels) components of an IDC. (c) RT–qPCR analysis of ADAM23 mRNA expression in paired in situ and invasive components of four IDCs (IDC1–4). Expression levels relative to normal breast tissue and normalized to the hypoxanthine guanine phosphoribosyltransferase (HPRT) endogenous control gene are shown in (a and c). (d and e) Immunohistochemistry staining of ADAM23 protein was carried out using the polyclonal antibody anti-ADAM23 (HPA012130). (d) ADAM23 protein expression in mammary normal ducts, in situ and invasive tumor components of an IDC (IDC15 from patient no. 15). Bar represents 100 μm. (e) ADAM23-ITH is observed between topographically distinct invasive areas of an undifferentiated IDC. In detail, mosaic clusters of A23pos (green arrow) and A23neg (red arrow) tumor cells coexisting in close proximity within an invasive component. Bar represents 50μm. Oncogene (2014), 1 – 10

© 2014 Macmillan Publishers Limited

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

3

Figure 2. ADAM23-ITH promotes cell proliferation and tumor growth. (a) 435-A23pos, 435-A23neg or a 1:1 mixture of both (435-A23het) were cultured as MS for 4 days and were pulsed for 6 h with BrdU. After MS dissociation, cell proliferation was evaluated by flow cytometric analysis of BrdU labeling (n = 10 000 cells per MS, 8 MS per group). NS, nonsignificant, **P o0.01 using one-way analysis of variance (ANOVA). (b and c) Cells were plated at clonogenic density in 0.6% soft agar and left for 2 weeks to allow colony formation. (b) The number of colonies formed per well (n = 3 independent experiments) and (c) the average colony size (area, mm2) (n = 20 colonies per experiment) were determined under the light microscope. ***P o0.001 using Student’s t-test. (d and e) 435-A23pos, 435-A23neg or 435-A23het mixtures were inoculated subcutaneously (s.c.) in female nude mice. (d) Kaplan–Meier curves depict s.c. tumor-free mice. Arrow indicates reduced tumor latency of 435-A23het tumors. ***Po 0.001 using a Mantel–Cox log-rank test. (e) S.c. tumor growth was monitored weekly by measuring two perpendicular tumor diameters. Arrow indicates the early increase in growth of 435-A23het tumors. ***P o0.001 using one-way ANOVA. (f and g) Red fluorescent 435-A23pos cells (435-A23pos/DiIC18) were mixed at different ratios with unlabeled 435-A23neg, allowed to aggregate as MS and proliferation was evaluated as described above. (f) Proliferation rates of 435-A23pos/DiIC18 cells in 435-A23het-MS (n = 10 000 DiIC18labeled cells per MS, 6 MS per group). (g) Proliferation rates of 435-A23neg cells in 435-A23het-MS (n = 10 000 unlabeled cells per MS, 6 MS per group). NS, nonsignificant, *Po 0.05 and ***P o0.001 using one-way ANOVA. Error bars indicate s.e.m. in all experiments.

clonogenic potential but distinct proliferative capacity in vitro (Figures 2b and c, P o 0.001). Accordingly, ADAM23 overexpression in CMS5a cells significantly promoted colony growth in anchorage-independent clonogenic assays (Supplementary Figure 3). In subcutaneous tumorigenesis assays, 435-A23neg cells showed impaired tumorigenesis (5/15, 33%) when compared with 435-A23pos (14/15, 93%, P o0.01) and 435-A23het (10/11, 91%, P o0.01) cells (Figure 2d). Significant differences in tumor growth kinetics and final tumor sizes were also observed between 435A23neg and the other groups (Figure 2e, P o 0.001). Interestingly, histological analysis of 435-A23neg injection sites from tumor-free mice revealed the presence of a small group of living tumor cells for periods up to 10 weeks after injection. These ADAM23negative ‘dormant’ tumors exhibited a late and slow growth 12 weeks after inoculation, and the real-time polymerase chain reaction (PCR) analysis revealed that ADAM23 silencing was stably maintained in these tumors during late in vivo growth (data not shown). It is worth noting that animals injected with 435-A23het mixtures exhibited reduced tumor latency than animals injected with either 435-A23pos or 435-A23neg cells (Figure 2d) and, in vitro, 435-A23hetMS had similar percentages of BrdU-labeled cells as 435-A23posMS, despite having initially half the amount of 435-A23pos cells (Figure 2a). Flow cytometry analysis of tumors from mice injected with 435-A23het mixtures revealed that 435-A23neg cells were retained in these tumors 10 weeks after inoculation, despite their proliferation deficiency (data not shown). Similar results were observed by immunohistochemical analysis of ADAM23 © 2014 Macmillan Publishers Limited

expression in 435-A23het tumors (Supplementary Figure 4). These results indicate that the coexistence of A23neg and A23pos cells in primary tumors enhance proliferation and tumor growth. To explore this possibility, we produced 435-A23het-MS composed of 435-A23pos cells prelabeled with a red fluorescent dye (435-A23pos/DiIC18) and different ratios of unlabeled 435A23neg cells, so that we could track the proliferation rate of both cell subpopulations in heterotypic MS. In 435-A23het-MS, the percentage of BrdU-labeled 435-A23pos/DiIC18 cells increased from 28 to 45% along with the increase in the number of 435-A23neg cells (Figure 2f, P o0.001), indicating that, in heterotypic environments, 435-A23neg cells enhance the proliferation rate of 435A23pos cells. In contrast, rates of BrdU labeling in 435-A23neg cells progressively decreased from 20 to 10%, with the gradual increase in 435-A23pos/DiIC18 cell number (Figure 2g, P o 0.05). To demonstrate that ADAM23-dependent proliferative signals are not restricted to MDA-MB-435 cancer cells, we screened a panel of tumor cell lines for ADAM23 mRNA expression (Supplementary Figure 5) and performed similar experiments with SK-Mel-37 melanoma cells (SK-A23pos) and their ADAM23shRNA derivatives (SK-A23neg) (Supplementary Figures 2c and d). In vitro, SK-A23neg-MS showed a smaller percentage of BrdUlabeled cells than SK-A23pos-MS (Supplementary Figure 6a). In vivo, subcutaneous tumorigenicity of SK-A23neg cells (5/7) was not impaired compared with SK-A23pos cells (5/7), although significant differences in growth kinetics and tumor size were observed (Supplementary Figure 6b, Po 0.01), Notably, SK-A23het mixtures were more tumorigenic (8/8) and showed reduced tumor latency Oncogene (2014), 1 – 10

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

4 compared with both homotypic groups (Supplementary Figure 6c, Po 0.05). Altogether, our findings demonstrate that ADAM23 expression is important for tumor growth at the primary site and that ADAM23 silencing is associated with an accentuated deficiency in cell proliferation, but not with reduced clonogenic cell survival. Moreover, we demonstrated that in ADAM23-heterotypic environments, the proliferation of the A23pos subpopulation is enhanced by A23neg cells, compensating for the intrinsic proliferative deficiency of the latter subpopulation. Most importantly, our results suggest that ADAM23-ITH can promote tumor growth. LGI4 mediates the proliferative crosstalk in ADAM23heterotypicenvironments LGI4 (also known as LGIL3), a member of the LGI family, is a putative secreted protein34 involved in proliferation in the nervous system35 and a known ADAM11, ADAM22 and ADAM23 ligand.35–37 We found that LGI4 mRNA expression was eightfold higher in 435A23neg than in 435-A23pos cells (Figure 3a, Po0.01). To determine whether LGI4 produced by 435-A23neg cells is responsible for the enhanced cellular proliferation of 435-A23pos cells in 435-A23het-MS, we examined BrdU labeling in A23het-MS composed of 435-A23neg

cells with or without ablation of LGI4 (435-A23neg/shLGI4 or 435A23neg/shCtrl, respectively) (Figure 3b). Consistent with a proproliferative function of LGI4, the percentage of BrdU labeling in 435-A23neg/shLGI4-MS (20%) was significantly lower than in 435-A23neg/shCtrl-MS cells (28%) and 435-A23pos-MS (34%) (Figure 3c, P o 0.001). Moreover, the 435-A23pos/DiIC18 subpopulation showed an increased percentage of BrdU labeling, from 32 to 40%, when cocultured with 435-A23neg/shCtrl cells (P o0.001), but not when cocultured with 435-A23neg/shLGI4 cells (31%, Figure 3d). As expected, LGI4 knockdown in 435-A23neg cells significantly impaired by 43% the previously observed enhanced growth of A23het subcutaneous tumors (Figure 3e). In agreement with an ADAM23-mediated proliferative crosstalk, LGI4-dependent tumor growth was exclusively determined by the coexistence of A23neg and A23pos cells, as subcutaneous inoculums of 435-A23neg/shCtrl or 435-A23neg/shLGI4 alone were essentially non-tumorigenic, because of the absence of ADAM23 expression. To our knowledge, this is the first demonstration of the relevance of LGI4/ADAM23 interactions for tumor cell proliferation and tumorigenesis. Moreover, our results show that LGI4 is the key factor mediating the proliferative crosstalk between A23neg and A23pos cells in ADAM23-heterotypic environments.

Figure 3. LGI4 mediates the proliferative crosstalk in ADAM23-heterotypic environments. (a) RT–qPCR analysis of LGI4 expression in 435A23pos and 435-A23neg cells. Expression levels relative to MDA-MB-435 wild-type cells and normalized to hydroxymethylbilane synthase (HMBS) endogenous control gene are shown (n = 4). **Po0.01 using Student’s t-test. (b) 435-A23neg cells were stably transduced with two different LGI4 shRNAs sequences (shLGI4 nos. 1 or 2) or with control shRNA (shCtrl) and LGI4 mRNA expression levels relative to cells transfected with shCtrl were determined as described above (n = 3). ***Po 0.001 using one-way analysis of variance (ANOVA). (c) 435-A23pos and 435-A23neg cells transfected with shLGI4 (A23neg/shLGI4) or shCtrl (A23neg/shCtrl) were allowed to aggregate as MS and proliferation was evaluated as described above (n = 10 000 cells per MS, 6 MS per group). ***P o0.001 using one-way ANOVA. (d) Proliferation of 435-A23pos/ DiIC18 cells in 435-A23het-MS composed of 90% of A23neg/shCtrl or A23neg/shLGI4 cells was determined as described above (n = 10 000 cells per MS, 4 MS per group). NS, nonsignificant, ***Po 0.001 using one-way ANOVA. (e) Subcutaneous A23het tumor growth constituted of 8 × 105 435-A23neg cells stably transduced with two LGI4 shRNAs sequences (grouped as shLGI4) or with shCtrl and 2 × 105 435-A23pos cells. A23homotypic inoculums (8 × 105 435-A23neg or 2 × 105 435-A23pos cells) were included as controls. **Po0.01 using one-way ANOVA. Error bars indicate s.e.m. in all experiments. Oncogene (2014), 1 – 10

© 2014 Macmillan Publishers Limited

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

5 ADAM23-ITH promotes tumor cell invasion and metastasis We next examined the role of ADAM23-ITH in tumor cell invasion and metastasis using MS embedded in collagen matrices, as well as spontaneous metastasis assays. Ten days after embedment in collagen matrices, 435-A23neg-MS were visibly more invasive than 435-A23pos-MS (Figure 4a). The average speed of invasion in 435-A23neg-MS was up to four times higher than in 435-A23pos-MS (4.7 and 1.6 μm/h, respectively, P o0.01, Figure 4b). It is worth noting that an invasive phenotype was accomplished in the noninvasive SK-Mel-37 cell line after ADAM23 ablation. The average invasion speed was 4.5 times higher (P o 0.01) for SK-A23neg-MS than for SK-A23pos-MS (Supplementary Figure 7). Accordingly, ADAM23 overexpression in CMS5a cells significantly inhibited invasion in collagen matrices (Supplementary Figure 8). We also carried out spontaneous metastasis assays by inoculating 435-A23pos, 435-A23neg and a 1:1 mixture of both (435-A23het)

into the mammary fat pad of immunodeficient mice. As expected, the percentage of animals that developed tumors was lower in mice inoculated with 435-A23neg cells (9/12, 75%) than with 435-A23pos cells (14/14, 100%, P o0.05, Figure 4c), but no significant differences in tumor growth, final sizes or tumor latency were observed between the groups when inoculated with the Matrigel into the mammary fat pad (Figures 4c and d). Unexpectedly, as ADAM23 silencing has been previously associated with metastasis development and poor prognosis,27 we observed no significant differences in the number of animals that developed metastases between the 435-A23neg (2/12, 16%) and 435-A23pos (5/14, 36%, P>0.05) groups (Table 1). As observed for the subcutaneous tumorigenesis assays, ADAM23 silencing remained stable throughout the spontaneous metastasis assay and, interestingly, metastatic lesions from animals bearing 435-A23het tumors displayed either a positive or negative pattern of ADAM23 mRNA expression, probably associated with the

Figure 4. ADAM23-ITH promotes cell invasion and metastasis. (a) Representative images of 435-A23pos-MS, 435-A23neg-MS and 435-A23het-MS embedded in collagen matrices for 10 days. (b) Invasion speed of MS presented in (a) (n = 4 MS per group). **P o0.01, ***Po 0.001 using oneway analysis of variance (ANOVA). (c and d) 435-A23pos, 435-A23neg or a 1:1 mixtures (435-A23het) were inoculated into the mammary fat pad (m.f.p.) of female nude mice. (c) Kaplan–Meier curves depict m.f.p. tumor-free mice. *P o0.05 using a Mantel–Cox log-rank test. (d) Tumor growth in the m.f.p. by measuring two perpendicular tumor diameters. (e) Invasion speed of 435-A23het-MS as a function of the percentage of 435-A23neg cells in the MS. *P o0.05, ***P o0.001 using one-way ANOVA. (f) Representative images of the invasiveness of 435-A23pos/GFP cells in 435-A23pos-MS (top panel) or in 435-A23het-MS (bottom panel). Red arrows highlight the distances traveled by 435-A23pos/GFP cells. The graph below represents the invasion speed of 435-A23pos/GFP cells in both environments (n = 4 MS per group). **P o0.01 using Student’s t-test. (g) Invasive bundles were formed by 435-A23pos/GFP and 435-A23neg/RFP. Invasive cells were outlined and reconstructed to compose a merged picture (detail). (h) Total number of migrating cells when 435-A23pos or 435-A23neg cells or 435-A23het mixtures were plated in the upper chamber of transwell (n = 4). ***P o0.001 using one-way ANOVA. (i) Number of 435-A23pos/GFP migrating cells when cocultured in the presence of either 435-A23pos or 435-A23neg cells (n = 3). *P o0.05 using Student’s t-test. (j) Number of 435-A23pos/GFP migration cells toward serum-free 24h conditioned medium (24 h-CM) from 435-A23pos or 435-A23neg cells (n = 3). ***P o0.001 using one-way ANOVA. Error bars indicate s.e.m. in all experiments. NS, nonsignificant. © 2014 Macmillan Publishers Limited

Oncogene (2014), 1 – 10

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

6 Table 1.

Metastatic potential of 435-A23pos, 435-A23neg and het 435-A23 tumors Mice with lung metastasis neg

435-A23 435-A23pos 435-A23het

2/12 (16%) 5/14 (36%)NS 11/23 (48%)*

Abbreviation: NS, nonsignificant. 435-A23pos or 435-A23neg cells or a 1:1 mixture of both (435-A23het) were suspended in the Matrigel and injected into the mammary fat pad of female Balb/c nude mice (106 cells in 0.05 ml of the Matrigel). Primary tumors were surgically removed and, after 8 weeks, animals were killed and the number of metastasis in lungs, kidneys and bladders were assessed by macroscopic observation at × 10 magnification. *P ⩽ 0.05 using a χ2 test.

different ADAM23 expression status of their founder’s tumor cells (Supplementary Figure 9). It is worth noting that the metastatic potential of 435-A23het tumors was threefold higher than that of tumors composed exclusively of 435-A23neg cells (11/23, 48%, P ⩽ 0.05) (Table 1) and, in vitro, 435-A23het-MS had a similar invasive behavior to 435A23neg-MS, despite having initially half the amount of 435-A23neg cells (Figure 4b, P o 0.001). In addition, we observed a positive correlation between the percentage of A23neg cells within the MS and the average invasion speed (Figure 4e, r = 0.67), indicating that ADAM23-ITH can enhance tumor invasion and metastasis. To explore this possibility, we produced A23het-MS, composed of GFP-labeled 435-A23pos cells (435-A23pos/GFP) and RFP-labeled 435-A23neg cells (435-A23neg/RFP), so that we could track the distance traveled daily over 12 days by both cell subpopulations in the heterotypic spheroids. We observed an increase in 60% in the invasion speed of 435-A23pos/GFP cells in A23het-MS compared with that in A23pos-MS (composed of labeled and unlabeled 435A23pos cells) (2.5 and 1.5 μm/h, respectively; Figure 4f, P o 0.01). No difference in the invasion speed of 435-A23neg/RFP cells within A23het-MS or A23neg-MS was observed (data not shown). Consistent with a cooperative invasive pattern, we frequently noticed interlaced subpopulations of 435-A23pos/GFP and 435-A23neg/RFP cells within single-cell bundles from A23het-MS (Figure 4g). Next, we performed haptotactic migration assays to determine whether the increased invasive behavior of A23pos cells is strictly dependent on the generation of a less resistant migratory pathway by A23neg cells in collagen matrices, or if A23neg cells also actively influence the migratory behavior of A23pos cells through paracrine signals. We observed that the number of migrating cells was higher in 435-A23het (250 cells per mm2) than in 435-A23pos cultures (22 cells per mm2, P o 0.001), but no significant differences were observed with 435-A23neg (205 cells per mm2, P > 0.05; Figure 4h). Considering that A23het mixtures contained half the number of A23neg cells, these results confirm that when mixed A23pos and A23neg cells have a different migratory behavior. We then carried out haptotactic assays to analyze the migratory behavior of 435-A23pos/GFP cells when mixed with unlabeled 435A23pos or 435-A23neg cells in the upper chamber of transwell plates. Notably, a 77% increase in the number of 435-A23pos/GFP migrating cells was observed when they were mixed with 435A23neg cells (from 40 to 77 cells per mm2; Figure 4i, Po0.05). Finally, we compared the chemotactic activity of 24 h conditioned medium obtained from 435-A23pos and 435-A23neg cells. We observed no differences in the chemotactic response of 435-A23pos cells to both conditioned medium, but a threefold increase in the number of 435-A23pos migrating cells with both conditioned medium compared with control medium (Figure 4j, Po0.001), suggesting that the enhancing effect of 435-A23neg cells on the migratory behavior of 435-A23pos cells depends on close proximity. Oncogene (2014), 1 – 10

Altogether, our findings demonstrate that A23neg cells are highly invasive, but not sufficient to seed metastasis and that, in an ADAM23-heterotypic environment, the invasive and migratory behavior of A23pos cells is enhanced by close proximity to A23neg cells. Most importantly, our results suggest that ADAM23-ITH positively affects tumor cell motility, invasion and, consequently, metastasis. NO mediates the invasive crosstalk in ADAM23-heterotypic environments NO is a transient species that exerts its physiological effects at short distances (up to 50 cell diameters).38 Considering the close proximity dependence of ADAM23-ITH for the invasive crosstalk and the important role of NO in tumor metastasis by promoting cell migration and invasion,39–41 we analyzed the expression of NO synthases (NOS1, NOS2 and NOS3) in A23pos and A23neg cells. We found that NOS3 expression (also known as eNOS) was upregulated in 435-A23neg cells compared with 435-A23pos cells (Supplementary Figure 10, P o 0.01), increasing NO levels by 56% (Figures 5a and b, P o 0.01). Similarly, SK-A23neg cells produced threefold more NO than SK-A23pos cells (Supplementary Figure 11, Po 0.001). To examine the effect of exogenous NO on the migratory phenotype of A23pos cells, we performed haptotactic assays in the presence of the slow-releasing NO donor SNAP (Snitroso-N-acetylpenicillamine). A 25% significant increase in the number of 435-A23pos migrating cells was observed at low SNAP concentrations (Figure 5c, P o 0.01). In agreement with this, pharmacological inhibition of NOS with L-NAME (L-NG-nitroarginine methyl ester) significantly inhibited the migration of 435A23pos cells in the presence of 435-A23neg cells, confirming that NO is a key mediator of ADAM23-mediated cell migration (Supplementary Figure 12). To investigate the effect of NO secreted by A23neg cells on the invasive behavior of A23pos cells, we ablated NOS3 expression in 435-A23neg cells using NOS3 shRNAs (435-A23neg/shNOS3; Figure 5d). Whereas the 30% reduction in endogenous NO production (Figure 5e, Po 0.01) did not alter the intrinsic invasive behavior of 435-A23neg cells (data not shown), it was sufficient to ablate completely the influence of 435-A23neg cells on the invasive behavior of 435-A23pos cells in A23het-MS (Figures 5f and g, Po 0.01). Consistently, the transwell migration of 435-A23pos/GFP cells increased by 60% when they were cocultured with 435A23neg/shCtrl (39 cells per mm2) compared with 435-A23neg/shNOS3 cells (24 cells per mm2; Figure 5h, P o0.01). These results demonstrate that NO production is upregulated in A23neg cells and NO is a key messenger mediating the invasive crosstalk between 435-A23pos and 435-A23neg cells DISCUSSION Here we showed, for the first time, that ADAM23 silencing can elicit a phenotypic switch from a proliferative to an invasive phenotype in tumor cells. Whereas A23pos tumor cells exhibit a fast-proliferating phenotype, high tumorigenicity and a lower invasive potential, A23neg tumor cells display a highly invasive phenotype, but an increased susceptibility to the trophic effect of the environment for optimal tumor growth. We also demonstrated, at both the mRNA and protein level, that ADAM23 expression is downregulated during breast cancer progression and ADAM23-ITH is a common characteristic of invasive breast carcinomas. ADAM23-ITH was observed in topographically distinct invasive areas of undifferentiated IDCs, with invasive components being frequently composed by mosaic clusters of A23pos tumor cells coexisting in close proximity with A23neg cells. Surprisingly, however, we observed that A23neg cells are not sufficient to seed metastasis owing to a severe deficiency to proliferate. Instead, the rise of A23neg cells within an © 2014 Macmillan Publishers Limited

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

7

Figure 5. NO mediates cooperative invasion in ADAM23-heterotypic environments. Quantification of NO production was performed using 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (DAF-FM) as described in Materials and methods. (a) Representative histograms of the quantification of NO production in 435-A23pos and 435-A23neg cells. The gray histogram represents the background fluorescence of cells without DAF-FM. (b) NO production of 435-A23pos and 435-A23neg cells (n = 9 per group). **P o0.01 using Student’s t-test. (c) Migration assays were carried out in the presence of the increasing concentrations of the slow-releasing NO donor S-nitroso-N-acetylpenicillamine (SNAP) (n = 2 per SNAP concentration). **Po 0.01 using one-way analysis of variance (ANOVA). (d) 435-A23neg cells were stably transduced with two different NOS3 shRNAs sequences (shNOS3 nos. 1 or 2) or with control shRNA (shCtrl) and NOS3 mRNA expression levels relative to cells transfected with shCtrl and normalized to hydroxymethylbilane synthase (HMBS) endogenous control gene were determined by RT–qPCR (n = 3). *Po0.05 using one-way ANOVA. (e) NO production in 435-A23neg/shNOS3 or 435-A23neg/shCtrl cells was measured as described above (n = 4 per group). **Po0.01 using Student’s t-test. (f) Representatives images of the invasive behavior of 435-A23pos/GFP cells in 435-A23pos-MS or in 435-A23het-MS composed of either 435-A23neg/shCtrl or 435-A23neg/shNOS3 cells. Red arrows highlight the distances traveled by 435-A23pos/ GFP cells in different environments. (g) Invasion speed of 435-A23pos/GFP cells in 435-A23pos-MS, 435-A23neg/shCtrl-MS or 435-A23neg/shNOS3-MS (n = 6 MS per group). **P o0.01 and ***P o0.001 using one-way ANOVA. (h) Number of 435-A23pos/GFP migrating cells when cocultured in the presence of either 435-A23neg/shCtrl or 435-A23neg/shNOS3 unlabeled cells (n = 5). **P o0.01 using Student’s t-test. Error bars indicate s.e.m. in all experiments. NS, nonsignificant.

ADAM23-positive tumor generates ADAM23-ITH that drives tumor growth and metastasis by actively enhancing the proliferative and invasive behavior of adjacent highly tumorigenic A23pos cells. Moreover, we demonstrated that LGI4 and NO are the key factors produced by A23neg tumor cells enhancing proliferation and invasion, respectively, of adjacent A23pos cells in heterotypic tumors. It has been proposed that ITH may drive tumor progression by creating an ecological niche in which tumor cells with different genotypes interact through paracrine signals and cooperate to promote tumor growth and dissemination.7 Identifying the key molecules mediating the cellular cooperation is thus promising avenues to reveal new approaches for the effective treatment of metastatic disease. Interactions between LGI and ADAM proteins are known to promote proliferation and differentiation of neuronal precursors cells and adipocytes.42,43 However, the findings reported here indicate that LGI/ADAM interactions are also relevant for tumorigenesis and ablation of LGI4 production in A23neg cells significantly attenuates proliferation and tumor growth in the A23-heterotypic environments. Curiously, changing the ratio of A23pos and A23neg cell subpopulations in A23het-MS significantly altered the proliferative differences between them. One explanation for these observations is that both subpopulations are competing for LGI4dependent proliferative signals, and ADAM23 expression confers an inherent advantage in this competitive environment. Moreover, considering that MDA-MB-435 cells do not express ADAM11 mRNA, but do express ADAM22 mRNA (Supplementary Figure 13), we speculate that the residual LGI4-dependent proliferation of 435-A23neg cells is supported by LGI4/ADAM22 interactions. © 2014 Macmillan Publishers Limited

The potential of NOS3 as a therapeutic target has recently been reconsidered44 and could also be particularly useful in the context of an ADAM23-heterotypic environment. Unusual NO production has been detected in different types of tumors, altering the cell redox state and signaling. Tumor-derived NO has been shown to enhance migration and invasion of cancer cells.39–41 Notably, NOS3-ITH has been observed in primary breast tumors, and higher levels of NOS3 were detected in undifferentiated tumor zones than in differentiated areas.40 In conclusion, based on our previous findings26,27 and on the work described here, we suggest that ADAM23 silencing, observed during breast cancer progression as a result of tumor epigenetic instability, generates a subpopulation of ADAM23negative clones within ADAM23-positive tumors and that metastatic spread and malignancy is in fact driven by the coexistence and interaction between ADAM23-negative and -positive subpopulations in the primary tumor. In this scenario, it is tempting to speculate that the previously observed association between ADAM23 hypermethylation and poor clinical outcome27 is indeed associated with ADAM23-ITH and to propose that additional studies addressing the association between ADAM23-ITH in primary tumors, metastatic disease and clinical outcome are worth pursuing. MATERIALS AND METHODS Laser microdissection Tumor cells were laser captured using the Pix Cell-II LCM System (ArcturusEngineering Inc., Mountain View, CA, USA) as described previously.45 Oncogene (2014), 1 – 10

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

8 Samples were free of stromal cells contamination. Breast DCIS samples are composed of tumor cells captured from ducts of pure DCIS lesion. IDC samples were divided geographically into two components: in situ and invasive components. In situ components are composed of tumor cells captured from mammary ducts and invasive components are composed of infiltrative tumor cells captured outside from the ducts. These experiments were conducted according to ethic standards and approved by the Ethics Committee in Human Research from FAP-AC Camargo (1143/08).

Reverse transcription and quantitative PCR Total RNA was extracted using Trizol (Life Technologies, New York, NY, USA). Reverse transcription (RT) was carried out with 2 μg DNA-free total RNA using Superscript II reverse transcriptase (Life Technologies). Quantitative PCR (qPCR) was performed using a 7300 Real-time PCR System and SYBR Green (Life Technologies). Relative expression levels were calculated as described.46 Hydroxymethylbilane synthase or hypoxanthine guanine phosphoribosyltransferase were used as endogenous control genes for expression data normalization.

Sodium dodecyl sulfate–polyacrylamide gel electrophoresis/ western blot Quantified membrane extracts were resolved on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis gel, transferred to polyvinylidene fluoride membrane (GE Healthcare Biosciences, Pittsburgh, PA, USA) and probed with antibody to ADAM23 (HPA012130; Sigma-Aldrich Co.) or to αv-integrin (AB1930; Millipore Co., Billerica, MA, USA) antibodies. Membranes were incubated for 60 min at room temperature in 0.1% Tween-20 with the secondary HRPT-conjugated antibody. Reactive proteins were visualized with the ECL Western Blotting Detection System (GE Healthcare Bioscience) according to the manufacturer's recommendations.

Generation of MS Semiconfluent monolayers cultures were suspended in the complete medium and 4 × 104 cells per well were plated in 48-well plates (Corning Life Sciences, Tewksbury, MA, USA) coated with 1% agarose in RPMI. After 4 days, a single MS spontaneously aggregated in each well. MS were readily handled with a sterile Pasteur pipette for subsequent experiments.

Immunohistochemistry

Proliferation rates by BrdU nuclear labeling

Frozen sections of tumor biopsies were fixed for 30 min in 4% paraformaldehyde, followed by 5 min denaturation in cold acetone. Endogenous peroxidase activity was depleted by incubation with 3% hydrogen peroxide solution. Sections were permeabilized for 1 h with 0.1% Triton X-100 and incubated overnight at 4 °C with anti-ADAM23 (HPA012130; Sigma-Aldrich Co., St Louis, MO, USA). ADAM23 immunostaining were visualized using a Tyramide Amplification Kit (T20921; Life Technologies) and 3,3′-diaminobenzidine substrate (Vector Laboratories, Burlingame, CA, USA). The tissue sections were counterstained with hematoxylin, dehydrated and coverslipped with Permount (Fisher Scientific, Waltham, MA, USA).

Four-day-old MS were incubated for 6 h at 37 °C in a complete medium containing 10 μM of BrdU (Pharmingen, San Diego, CA, USA). After this period, MS were dissociated with Trypsin-EDTA and single-cell suspensions were fixed, permeabilized and incubated with fluorescein isothiocyanateconjugated anti-BrdU antibody in accordance with the manufacturer's protocol (BD Pharmingen). BrdU-labeled nuclei were analyzed in 104 cells per sample on a FACSCalibur cytometer (BD Pharmingen).

Cell lines and cell culture The MDA-MB-435 tumor cell line was obtained from the ATCC (Manassas, VA, USA), the SK-Mel-37 human melanoma and CMS5a murine sarcoma cell lines were donated by Dr LJ Old and by Dr G Ritter (Memorial SloanKettering Cancer Centre, New York, NY, USA). All cells were maintained at 37 °C, under 5% CO2, in a complete medium: RPMI 1640 (Life Technologies) supplemented with 10% fetal bovine serum (Cultilab, Campinas, SP, Brazil) and 1% L-glutamine (Sigma-Aldrich Co.).

Ablation of ADAM23, LGI4 and NOS3 genes by shRNAs and ADAM23 overexpression shRNA ADAM23 ablation was performed as described previously.27 For the sake of simplicity, results obtained for the wild-type cells and cells transfected with a control shRNA vector were grouped throughout the text and presented as the A23pos, whereas results obtained for two independent ADAM23-shRNA clones (shA23 nos. 1 and 2) were grouped and presented as A23neg. LGI4 (shLGI4) and NOS3 (shNOS3) ablations were obtained by lentiviral infection using pLKO.1 vector containing two different shRNA sequences obtained from Sigma-Aldrich Co. shRNAs to GFP sequence was used as control (shCtrl) (Sigma-Aldrich Co.). Viral production was carried out using Lipofectamine-mediated transfection of human embryonic kidney 293T cells. CMS5a cells were transfected with pcDNA3.1 empty vector or pcDNA3.1-ADAM23-HA (splicing isoform alpha) kindly donated by Dr S Cal.47

Membrane isolation protocol A total of 5 × 106 cells were washed with ice-cold phosphate-buffered saline and scraped into the membrane extraction buffer (50 mM Tris–HCl (pH 7.4), 10% sucrose and 5 mM EDTA) containing protease inhibitors and were incubated at 4 °C under agitation for 30 min. Cells were disrupted by passing through a 27G needle and cell membranes were precipitated by centrifugation at 18 000 g for 75 min. Pellets were suspended in NP40 buffer (20 mM Tris–HCl (pH 8.0), 137 mM NaCl, 10% glycerol, 1% Nonidet P-40 and 2 mM EDTA) containing protease inhibitors for 6 h and cleaned at 18 000 g for 30 min. Oncogene (2014), 1 – 10

Clonogenic anchorage-independent assays Single-cell suspensions were mixed to 500 μl of 0.6% agarose diluted in the complete medium and were plated in culture dishes coated with 5 mm of 1% agarose. Plates were incubated at 37 °C and 5% CO2 for 2 weeks to allow colony formation. The number of colonies formed and the colony area (mm2) were determined under the light microscope using the ImageJ software (NIH, Bethesda, MD, USA).

Subcutaneous tumorigenesis assays In all, 106 tumor cells suspended in 100 μl of phosphate-buffered saline were inoculated subcutaneously in the flank of immunodeficient female BALB/c nude mice (4–10 weeks old). Tumor growth was monitored weekly and tumor sizes were determined by measuring two perpendicular tumor diameters through the tumor long axis using a caliper. All experiments were conducted according to ethic standards and approved by the Animal Experimentation Ethics Committee of the Medical and Research Centre at Hospital AC Camargo (043/09).

Invasion assays into three-dimensional collagen matrices MS were embedded in 1.5–3 mg/ml of reconstituted rat tail type-1 collagen (BD Pharmingen) and plated in 48-well plates coated with 5 mm of 1% agarose. After polymerization, collagen matrices were overlaid with the complete medium. Average invasion speed was obtained dividing the distances, from the MS border, traveled by ‘leader’ cells (80–100 tumor cells per spheroid) by time. Invasion was determined under the bright field or fluorescence microscopy using the ImageJ software (NIH).

Spontaneous metastasis assay Metastatic assays were performed as described previously.48 ADAM23heterotypic tumors were generated by inoculation of a single-cell suspension of A23pos and A23neg cells at a 1:1 ratio. Primary tumors were surgically removed when they reached a mean diameter of 1 cm tumors and skin incisions were closed. After 8-week interval, animals were killed and autopsied. These experiments were conducted according to ethic standards and approved by the Animal Experimentation Ethics Committee from FAP– AC Camargo (043/09).

Haptotactic migration assay Migration assays were carried out using transwell plates (Corning Life Sciences) with 8 μm polycarbonate filters precoated with 20 μg/ml rat tail © 2014 Macmillan Publishers Limited

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

9 collagen type-1 (BD Pharmingen). A total of 4 × 104 cells were plated in the upper chamber of the transwell and allowed to migrate for 8 h at 37 °C. At the end of the experiment, the top side of the filters were scraped with cotton swabs and migrating cells were fixed and stained with 4′,6-diamidino-2-phenylindole (Sigma-Aldrich Co.). Cells were counted at x100 magnification in 20 different optical fields/insert.

Fluorometric NO quantification The cell-permeant fluorophore DAF-FM (4-amino-5-methylamino-2′,7′difluorofluorescein diacetate)49 was used in accordance with the manufacturer's protocol (Life Technologies). Fluorescence of 2 × 104 cells per sample was measured on a FACSCalibur cytometer (BD Pharmingen) and expressed as arbitrary fluorescence units.

CONFLICT OF INTEREST The authors declare no conflict of interest.

ACKNOWLEDGEMENTS We are grateful to Dr Andrew Simpson and Dr Raphael Parmigiani for critical reading of this manuscript and to Dr Rodrigo Perez, Dr Alicia Kowaltowski, Dr Beatriz Geronymo, Dr Camila Machado, Dr Vilma Martins, Dr Tiago dos Santos and Dr Anibal Vercesi for scientific discussions and technical assistance. This work was supported by grants from FAPESP (06/60219-2, 07/51194-9, 07/52920-5, 11/50541-2) and LICR.

REFERENCES 1 Cairns J. Mutation selection and the natural history of cancer. Nature 1975; 15: 197–200. 2 Fidler IJ, Kripke ML. Metastasis results from preexisting variant cells within a malignant tumor. Science 1977; 197: 893–895. 3 Marusyk A, Almendro V, Polyak K. Intratumor heterogeneity: a looking glass for cancer? Nat Rev Cancer 2012; 19: 323–334. 4 Merlo LM, Pepper JW, Reid BJ, Maley CC. Cancer as an evolutionary and ecological process. Nat Rev Cancer 2006; 6: 924–935. 5 Michor F, Polyak K. The origins and implications of intratumor heterogeneity. Cancer Prev Res 2010; 3: 1361–1364. 6 Nowell PC. The conal evolution of tumor cell populations. Science 1976; 194: 23–28. 7 Swanton C. Intratumor heterogeneity: evolution through space and time. Cancer Res 2012; 72: 4875–4882. 8 Suzuki N, Withers HR, Koehler MW. Heterogeneity and variability of artificial lung colony-forming ability among clones from mouse fibrosarcoma. Cancer Res 1978; 38: 3349–3351. 9 Anderson K, Lutz C, van Delft FW, Bateman CM, Guo Y, Colman SM et al. Genetic variegation of clonal architecture and propagating cells in leukaemia. Nature 2011; 469: 356–361. 10 Clark J, Attard G, Jhavar S, Flohr P, Reid A, De-Bono J et al. Complex patterns of ETS gene alteration arise during cancer development in the human prostate ETS gene rearrangements in prostate tissues. Oncogene 2008; 27: 1993–2003. 11 Cottu PH, Asselah J, Lae M, Pierga JY, Diéras V, Mignot L et al. Intratumoral heterogeneity of HER2/neu expression and its consequences for the management of advanced breast cancer. Ann Oncol 2008; 19: 595–597. 12 Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E et al. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. New Engl J Med 2012; 366: 883–892. 13 Maley CC, Galipeau PC, Finley JC, Wongsurawat VJ, Li X, Sanchez CA et al. Genetic clonal diversity predicts progression to esophageal adenocarcinoma. Nat Genet 2006; 38: 468–473. 14 Navin N, Kendall J, Troge J, Andrews P, Rodgers L, McIndoo J et al. Tumor evolution inferred by single-cell sequencing. Nature 2011; 7: 90–94. 15 Notta F, Mullighan CG, Wang JC, Poeppl A, Doulatov S, Phillips LA et al. Evolution of human BCR–ABL1 lymphoblastic leukemia initiating cells. Nature 2011; 469: 362–367. 16 Takahashi T, Habuchi T, Kakehi Y, Mitsumori K, Akao T, Terachi T et al. Clonal and chronological genetic analysis of multifocal cancers of the bladder and upper urinary tract. Cancer Res 1998; 15: 5835–5841. 17 Teixeira MR, Pandis N, Bardi G, Andersen JA, Heim S. Karyotypic comparisons of multiple tumorous and macroscopically normal surrounding tissue samples from patients with breast cancer. Cancer Res 1996; 56: 855–859.

© 2014 Macmillan Publishers Limited

18 Yachida S, Jones S, Bozic I, Antal T, Leary R, Fu B et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature 2010; 467: 1114–1119. 19 Yancovitz M, Litterman A, Yoon J, Ng E, Shapiro RL, Berman RS et al. Intra- and inter-tumor heterogeneity of BRAFV600E mutations in primary and metastatic melanoma. PLoS One 2012; 7: e29336. 20 Almendro V, Marusyk A, Polyak K. Cellular heterogeneity and molecular evolution in cancer. Annu Rev Pathol 2013; 24: 277–302. 21 Inda MM, Bonavia R, Mukasa A, Narita Y, Sah DW, Vandenberg S et al. Tumor heterogeneity is an active process maintained by a mutant EGFR-induced cytokine circuit in glioblastoma. Genes Dev 2010; 24: 1731–1745. 22 Yap TA, Gerlinger M, Futreal PA, Pusztai L, Swanton C. Intratumor heterogeneity: seeing the wood for the trees. Sci Transl Med 2012; 28: 127ps10. 23 Axelrod R, Axelrod DE, Pienta KJ. Evolution of cooperation among tumor cells. Proc Natl Acad Sci USA 2006; 103: 13474–13479. 24 Blobel CP. ADAMs: key components in EGFR signalling and development. Nat Rev Mol Cell Biol 2005; 6: 32–34. 25 Costa FF, Colin C, Shinjo SM, Zanata SM, Marie SK, Sogayar MC et al. ADAM23 methylation and expression analysis in brain tumors. Neurosci Lett 2005; 380: 260–264. 26 Costa FF, Verbisck NV, Salim AC, Ierardi DF, Pires LC, Sasahara RM et al. Epigenetic silencing of the adhesion molecule ADAM23 is highly frequent in breast tumors. Oncogene 2004; 23: 1481–1488. 27 Verbisck NV, Costa ET, Costa FF, Cavalher FP, Costa MD, Muras A et al. ADAM23 negatively modulates alpha(v)beta(3) integrin activation during metastasis. Cancer Res 2009; 69: 5546–5552. 28 Hagihara A, Miyamoto K, Furuta J, Hiraoka N, Wakazono K, Seki S et al. Identification of 27 5′-CpG islands aberrantly methylated and 13 genes silenced in human pancreatic cancers. Oncogene 2004; 23: 8705–8710. 29 Takada H, Imoto I, Tsuda H, Nakanishi Y, Ichikura T, Mochizuki H et al. ADAM23, a possible tumor suppressor gene, is frequently silenced in gastric cancers by homozygous deletion or aberrant promoter hypermethylation. Oncogene 2005; 24: 8051–8060. 30 Calmon MF, Colombo J, Carvalho F, Souza FP, Filho JF, Fukuyama EE, et al. Methylation profile of genes CDKN2A (p14 and p16), DAPK1, CDH1, and ADAM23 in head and neck cancer. Cancer Genet Cytogenet 2007; 173: 31–37. 31 Choi JS, Kim KH, Jeon YK, Kim SH, Jang SG, Ku JL et al. Promoter hypermethylation of the ADAM23 gene in colorectal cancer cell lines and cancer tissues. Int J Cancer 2009; 124: 1258–1262. 32 Hu C, Lv H, Pan G, Cao H, Deng Z, Hu C et al. The expression of ADAM23 and its correlation with promoter methylation in non-small-cell lung carcinoma. Int J Exp Pathol 2011; 92: 333–339. 33 Yamada KM, Cukierman E. Modeling tissue morphogenesis and cancer in 3D. Cell 2007; 130: 601–610. 34 Senechal KR, Thaller C, Noebels JL. ADPEAF mutations reduce levels of secreted LGI1, a putative tumor suppressor protein linked to epilepsy. Hum Mol Genet 2005; 14: 1613–1620. 35 Nishino J, Saunders TL, Sagane K, Morrison SJ. LGI4 promotes the proliferation and differentiation of glial lineage cells throughout the developingperipheral nervous system. J Neurosci 2010; 45: 15228–15240. 36 Sagane K, Ishihama Y, Sugimoto H. LGI1 and LGI4 bind to ADAM22, ADAM23 and ADAM11. Int J Biol Sci 2008; 4: 387–396. 37 Ozkaynak E, Abello G, Jaegle M, van Berge L, Hamer D, Kegel L et al. Adam22 is a major neuronal receptor for LGI4-mediated Schwann cell signaling. J Neurosci 2010; 30: 3857–38564. 38 Garthwaite J, Boulton CL. Nitric oxide signaling in the central nervous system. Annu Rev Physiol 1995; 57: 683–706. 39 Gauthier N, Lohm S, Touzery C, Chantôme A, Perette B, Reveneau S et al. Tumorderived and host-derived nitric oxide differentially regulate breast carcinoma metastasis to the lungs. Carcinogenesis 2004; 25: 1559–1565. 40 Jadeski LC, Hum KO, Chakraborty C, Lala PK. Nitric oxide promotes murine mammary tumor growth and metastasis by stimulating tumor cell migration, invasiveness and angiogenesis. Int J Cancer 2000; 86: 30–39. 41 Orucevic A, Bechberger J, Green AM, Shapiro RA, Billiar TR, Lala PK. Nitric-oxide production by murine mammary adenocarcinoma cells promotes tumor-cell invasiveness. Int J Cancer 1999; 81: 889–896. 42 Bermingham JR, Shearin H, Pennington J, O'Moore J, Jaegle M, Driegen S et al. The claw paw mutation reveals a role for LGI4 in peripheral nerve development. Nat Neurosci 2006; 9: 76–84. 43 Kim HA, Park WJ, Jeong HS, Lee HE, Lee SH, Kwon NS et al. Leucine-rich glioma inactivated 3 regulates adipogenesis through ADAM23. Biochim Biophys Acta 2012; 6: 914–922. 44 Lampson BL, Kendall SD, Ancrile BB, Morrison MM, Shealy MJ, Barrientos KS et al. Targeting eNOS in pancreatic cancer. Cancer Res 2012; 72: 4472–4482.

Oncogene (2014), 1 – 10

ADAM23-ITH promotes tumor growth and metastasis ET Costa et al

10 45 Sens-Abuázar C, Napolitano E, Ferreira E, Osório CA, Krepischi AC, Ricca TI et al. Down-regulation of ANAPC13 and CLTCL1: early events in the progression of preinvasive ductal carcinoma of the breast. Transl Oncol 2012; 5: 113–123. 46 Pfaffl MW. A new mathematical model for relative quantification in real-time RT–PCR. Nucleic Acids Res 2001; 29: e45. 47 Cal S, Freije JMP, López JM, Takada Y, López-Otín C. ADAM 23/MDC3, a human disintegrin that promotes cell adhesion via interaction with the αvβ3

integrin through an RGD-independent mechanism. Mol Biol Cell 2000; 11: 1457–1469. 48 Price JE, Polyzos A, Zhang RD, Daniels LM. Tumorigenicity and metastasis of human breast carcinoma cell lines in nude mice. Cancer Res 1990; 50: 717–721. 49 Kojima H, Nakatsubo N, Kikuchi K, Kawahara S, Kirino Y, Nagoshi H et al. Detection and imaging of nitric oxide with novel fluorescent indicators: diaminofluoresceins. Anal Chem 1998; 70: 2446–2453.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

Oncogene (2014), 1 – 10

© 2014 Macmillan Publishers Limited